Worked on backend development for the ahujasid/blender-mcp repository, focusing on enhancing security in Python-based automation tools. Addressed prompt-injection risks by removing vulnerable lines from the docstrings of key MCP tool functions, specifically targeting get_hyper3d_status and get_hunyuan3d_status. This change improved prompt hygiene and reduced the risk of hidden steering in large language model prompts, thereby strengthening the reliability of automated responses. The update did not affect user-facing interfaces but contributed to safer backend operations. Emphasized security best practices and ensured traceability by linking the changes to a specific commit and issue closure for future audits.